The resulting trees are not equivalent to the initial ones. All
offending branches were split and their non-contiguous part was moved
directly below the root.

The previous implementation was failing when nodes were numbered in
breadth-first order. Also gave a more descriptive name to the
algorithm.

Feature/swc parser whitespace

Whitespace-only lines are now treated correctly in SWC file parser. This
automatically leads to a consistent and nice handling of old mac-style
eol; they are just another whitespace (not newline).

Refined the test cases a bit.

Added support for mapping swc record types to nest::mc::segmentKind.
This required the adjustment in some tests + the ball_and_stick file, since
the dendrite type of swc record is type '3'. Before, everything except
soma was added (hardcoded) as a dendrite. An swc_parse_error will be
thrown in case of an impossible mapping. Also, no check is performed for
inconsistencies of SWC record types along a branch; the type of the
branch is assumed to be the type of the first record (compartment) in
the branch.

Finally, adapted to the coding style.

It was complaining for std::accumulate() in my laptop with clang.

Add save_getline with windows EOL support

Use library's own version of assert.

Unit tests for probe functionality

* Also exposes current values from fvm_cell by segment location.

Issue 34 parse errors when opening a file created in windows.

Solution is a method that reads the lines in a windows and linux save manner.
Includes a unit/regression tests toggling between red and green after the fix is applied.

Feature/spike handling

* change `probe_sort` enum to scoped enum
    - renamed to `probeKind`
    - refactored out of class
    - updated coding guidelines wiki with enum rules
* refactor `std::pair` into structs with meaningfull name types in `cell.hpp`
    - not just probes: stimulii and detectors too.
* add profiling region around sampling in `cell_group`
* change output data format for traces to json
* remove white space at end of lines (looking at you Sam)

Feature/probes

Fix linking of MPI libraries for tests and validation.

* Finds an actual (possibly) reasonable use for std::endl.

Sidesteps NDEBUG versus assert() issue; EXPECTS macros are
now enabled via a CMake configuration option -DWITH_ASSERTIONS=ON,
corresponding to the preprocessor define -DWITH_ASSERTIONS in
the code itself.

(Note, with assertions enabled, one of the existing tests in
test_algorithms.cpp aborts.)

- profiler generates one report per MPI rank
- estimates the thread imbalance per rank
- generates a json file with profiling stats for each rank
- works with/without MPI
- works with/without TBB

Conflicts:
	miniapp/miniapp.cpp

- move definition of `global_policy` to header file
  src/communication/global_policy.hpp
- the `#ifdef WITH_MPI` guards go there so that
    - they don't pollute user code
    - they are available throughout the application
- small cleanup of a couple of style points in the mpi code

* Make global_policy methods static
* Move global_policy init/finalize into static methods
* Make miniapp use only global_policy methods for
  setup, rank, etc.

* First implementation of simple current/voltage probe
  interface
* Sample a specific collection of voltage or current
  values at independent dt (current implementation
  chooses to provide at most one sample per simulation
  dt per request.)
* Miniapp uses samplers to write soma and dendrite
  membrane potentials from first three cells.
* Also: fix noncompliant use of nullptr -> bool conversion
  in test_optional.cpp

Feature/spike handling